Michael Hauschild (CERN), LCWS02, 27-Aug-2002, page 1
2D(3D?) cluster counting with GEMs and small
pads:the digital TPC?
or
how to measure dE/dx without measuring
charges
Michael Hauschild (CERN), LCWS02, 27-Aug-2002, page 2
Ideal dE/dx measurement
Count number of clusters along track
cluster density should be proportional to dE/dx
Obvious problem:
cluster density is high (20 - 30 clusters/cm in Ar mixtures for m.i.p.) = 1 cluster per 300 - 500 ìm
need device with high granularity to resolve them in space
Other problem:
Clusters sometimes have more than one electron: 1 el. 82.4 %
2 el. 6.9 %
3 el 2.0 %
10 el. 0.64 %
100 el. 0.0014 %
how to avoid counting individualelectrons of multi-electron clusters?
(TESLA-TPC, Ar/CH4/CO2, 93/5/2, calculation by
HEED)
Michael Hauschild (CERN), LCWS02, 27-Aug-2002, page 3
Classic dE/dx measurement
Widely used (because counting is difficult):
measure charge over some track length (sampling length)
"average" charge of many samples = dE/dx
Charge measurement requires:
electronics with good charge resolution, e.g. 8 bit or more
stable gain!!!
physics needs ÄG/G < 1/10 ó(dE/dx)/dE/dx (better < 1/20)
some algorithm to remove unwanted multi-electron clusters (delta electrons)
commonly used: truncated mean, remove a fixed fraction of highest charge measurements, typically 20-30%, robust
< 0.5% (better < 0.2%) overall gain stability
Michael Hauschild (CERN), LCWS02, 27-Aug-2002, page 4
dE/dx with GEMs
GEMs (Gas Electron Multiplier) + pads (typical size 5-10 mm) are suggested as possible TPC-readout device
Nice detector for tracking, what about dE/dx?
Problem: GEMs show gas gain variations
local variations over the surface (static)
could make calibrations more complicated, not a problem in principle
time dependent (dynamic) variations due to charge-up effects
difficult to control and to calibrate, depend on background, might vary within a bunch train
Michael Hauschild (CERN), LCWS02, 27-Aug-2002, page 5
GEM gain variations
local gain variations: dynamic gain variations:
M. Hamann et al.(DESY/Univ. Hamburg) C. Altunbas et al., CERN-EP 2002-008
COMPASS GEMs
➔ 10% local gain variations ➔ 20-30% dynamic gain variations
Michael Hauschild (CERN), LCWS02, 27-Aug-2002, page 6
Cluster CountingDirect cluster counting avoids any problems with gas gain instabilities
In theory ultimate way to get dE/dx
30 clusters/cm * 120 cm track length = 3600 clusters = 1.7% dE/dx resolution (TESLA-TDR: 4-4.5%)
Not a brand new idea:
previous attempts tried to resolve clusters in time:
slow gas / drift velocity (e.g. CO2) + good time and multi-hit
resolution, worked in lab + prototype detectors, never used in real big detectors for physics
Now (that's new):
micro-pattern devices + small pads = high granularity could make it possible to resolve them in space (2D), if time could be added even 3D(?)
need only 1-bit "ADC"
Michael Hauschild (CERN), LCWS02, 27-Aug-2002, page 7
Could it work?
Simulation study made:
generate clusters/electrons (also long range delta electrons using HEED (by I. Smirnov), take gas parameters (diffusion etc.) from MAGBOLTZ (by S. Biagi)
track electrons through TPC volume, squeeze them through GEM holes, apply gas gain (use Polya distribution for fluctuations)
track all electrons created in gas amplification to a pad plane (including diffusion, ódiffusion = 135 ìm over 2 mm)
collect electrons on pads, allow 5% losses in gaps, add noise (200 el. R.M.S. per pad, optimistic?)
apply threshold (1500 el.) and simply count number of pads above threshold = clusters(?)
➔ very CPU time consuming, need 50 Mill. electrons per full TPC track
Michael Hauschild (CERN), LCWS02, 27-Aug-2002, page 8
Generated Electrons
HEED calculation
GEM plane
Track(clusters)
some delta-electron
TPC frame(sideview)
Michael Hauschild (CERN), LCWS02, 27-Aug-2002, page 9
Generated Electrons (close view)
no diffusion
with diffusion
(ódiff,trans. = 1.1 mm, ódiff,long. = 4.4 mm for 250 cm drift and 4 T)
65 keVdelta-electron
Michael Hauschild (CERN), LCWS02, 27-Aug-2002, page 10
Pad view (xy plane)
multi-electron cluster
single electrons at GEM plane
pads (500 x 500 ìm2) above threshold (1500
e-)
Michael Hauschild (CERN), LCWS02, 27-Aug-2002, page 11
Questions
Counting with small pads seems to work
Some questions to answer:
optimal pad size?
large pads: clusters can't be resolved
small pads: (too) many channels (cost!), bad signal/noise ratio
noise?
need low threshold to count a single electron after gas amplification on a single small pad
diffusion?
at large drift length (up to 250 cm at TESLA-TPC), multi-electron clusters are spread by diffusion, individual electrons appear and are counted again, not clusters
Michael Hauschild (CERN), LCWS02, 27-Aug-2002, page 12
Counts vs Pad Size
m.i.p., 0.6 GeV pions
high E tracks, 1000 GeV pions
(Fermi plateau)
Michael Hauschild (CERN), LCWS02, 27-Aug-2002, page 13
Separation Power
simple pad counting,2.1 ó separation
classic dE/dx by charge measurement + truncated mean, 2.1 ó separation
Michael Hauschild (CERN), LCWS02, 27-Aug-2002, page 14
Conclusions ISeparation power with cluster counting as good as classic dE/dx by charge + truncated mean...
...but not better! (Factor 2 improvement expected)
Need to match: average distance between clusters
(375 ìm for m.i.p.)
diffusion spread(1.1 mm for 250 cm
drift)
pad size(noise cost)
➔ Doesn't fit together for TESLA-TPC, too much diffusion!
➔ Electrons from multi-electron clusters dissolve and are individually counted (instead of clusters)
Michael Hauschild (CERN), LCWS02, 27-Aug-2002, page 15
Conclusions II
Cluster counting better than classic dE/dx by charge needs:
gas with low ionisation
large distance between clusters
low probability for multi-electron clusters
low diffusion
either low diffusion gas or
short drift length
pad sizes
Helium(Neon?) mixtures could be a possible candidate for successful cluster counting